In the present work, a method for indirect determination of the weight of Japanese quail eggs is proposed, taking into account changes in their internal properties. Visual data and transmission spectra is used. Shape features and spectral indices are selected and applied. It has been found that egg weight M can be predicted by the volume V of eggs and the spectral index GLI, M=f(V,GLI). The resulting model has a coefficient of determination R2=0,89, low error values, up to 3%. Mean square error MSE=0,03 and root mean square error RMSE=0,2. The results obtained can be used to indirect determination the weight of Japanese quail eggs when incubated, packaged.

In the present work an analysis of the separability of hen egg yolks from different manufacturers is made using image and spectral processing and analysis techniques. Apparent properties of three types of egg yolks were determined and a comparative analysis of these properties was made. Discriminant and SVM (Support vector machines) classifiers were used for classification. A general classification error with lower values ​​is obtained with the b (Lab) color component. In the studies of the spectral characteristics of egg yolks from different manufacturers, the highest accuracy of separation of the target areas is obtained with the kernel SVM classifier combined with the kernel variant of the principal components. When using this classifier, general classification errors of up to 1% were obtained. The results confirm the hitherto known research in this area because the major part of the chicken egg yolk properties studied are visible properties that can be analyzed in the visible spectrum of the reflected light.

The report presents a comparative analysis of algorithms for counting objects in images. They are used in counting eggs. From the three algorithms compared Threshold, Circular Hough and Wateshed, with high performance and small error values ​​is the algorithm Circular Hough. When recognizing the eggs, it is necessary to make a selection of the color model, by which to separate the eggs from the background according to their color and the breed of birds. More research is needed on the impact of the image capturing conditions on the accuracy of algorithms.

In this article a comparative analysis is made to determine the influence of vectors of selected features derived from geometric, optical and dielectric characteristics of eggs on the accuracy of classification, depending on their weight. Suitable for classification are the principal components and latent variables that reduce feature vectors containing shape indices (D, A, V), spectral indices (TVI, GLI), dielectric characteristics (C, k), selected by four methods (CORR, SFCPP, RELIEFF, FSRNCA). By comparative studies it is found that the use of classification methods (DT, DA, SVM) are more effective in predicting weight of hen eggs than in quail eggs. The proposed egg analysis methods take precedence over the known solutions in this field as it takes into account changes in the internal properties of quail and hen eggs when stored.

No abstract