| It is one of difficult problems to be resolved in egg hatch industry to identify the fertile information of hatching eggs and eliminate infertile eggs prior to the incubation. Many infertile eggs were wasted in the process of incubation every year, which caused considerable economic loss. Therefore, the detection of infertile eggs prior to incubation can improve the economic efficiency of incubation and the quality of egg processing in late period, and it can bring considerable economic benefits. The existing domestic infertile egg detection mainly depends on traditional manual candle method. However, this detection method requires high intensity of labor and is time consuming. In addition, the result of detection is subjective and its accuracy can not be guaranteed. This paper is aimed to propose the application of transmission hyperspectral image technology combining image and spectrum information to detect the fertile information of hatching eggs before incubation, in order to improve the hatch rate of hatching eggs and the economic efficiency of incubation. This will provide a theoretical basis and technical support for the real-time online detection.This paper took the white shell Jingfen no.1 hatching eggs from Yukou poultry industry in Jingzhou city, and white Leghorn hatching eggs from Huazhong Agricultural University chicken farm as the research object, application of transmission hyperspectral image technology, the nondestructive identification method of fertile and infertile eggs prior to the incubation were studied. Specific work were as follows:(1) A hypersspectral transmission image acquisition system was built. The light source, light intensity, resolution, exposure time, the platform moving speed, and other parameters were adjusted when the images of hyperspectral instrument were captured. Ultimately, the exposure time of the camera was determined as 0.1s, the resolution of image as 400×400, hatching egg collection speed as 1.7mm / s. Before hatching eggs incubation, hyperspectral imaging system was used to acquire the images whose wavelength was between 400 nm and 1000 nm.(2) The four image characteristics(ratios of length to short axis, elongation, roundness, the ratios of the yolk area to the whole area) was extracted through imageprocessing and analysis of fertile and infertile eggs. ENVI software was used to extract 520 waveband spectral information of hatching eggs. The spectrum between 400 nm and 1000 nm were divided into three different spectral regions. They were visible light, near infrared and full band respectively. Ultimately, the visible light were chosen to classify actual type of hatching eggs. Different spectral pretreatment methods were used to analyze spectra, e.g. Multiplicative Scatter Correction, Normalize, Standard Normal Variate Transformation, First Derivative, MSC+FD, SNV+FD, Normalize + FD. Among these methods, the Normalized pretreatment method was the most effective whose classification accuracy was better than other methods. In consequence, the Normalized method was used to process spectra, and then 155 spectral characteristic variables were extracted through the Normalization method and correlation coefficient method. Principal component analysis(PCA) method was adopted to reduce the dimensions of combining 4 image characteristic variables and 155 spectral characteristic variables together fusion information, top 6 principal components were extracted in PCA method to reduce the dimensions of representative image-spectrum fusion information.(3) There were 300 samples of hatching eggs. In accordance with the distribution principle of 2:1, training set and testing set were built. Support vector machine(SVM) method and Relevance vector machine(RVM) were used to establish classification models, which were based on 4 image characteristic variables, one hundred and fifty five spectral characteristic variables and image-spectrum fusion information respectively. The identification accuracy rates of the models were 84%, 90%, 90%, 91%, and 93%, 96%, respectively.The experimental results showed that the model based on image-spectrum fusion information model was superior to the single information model. Using hyperspectral transmission imaging technology combing the image and spectrum information to detect the fertile information of hatching eggs before incubation, is effective and feasible. RVM model was superior to SVM model in detecting fertile information of hatching eggs before incubation hatching eggs. The classification accuracy of RVM reached 96%. The indentifying speed of fertile and infertile eggs were very quickly. This research would provide theoretical basis for the real-time online detection and testing instrument for hatching eggs. |
PDF Full Text Request