基于可见-近红外高光谱技术对鸡种蛋性别鉴定的研究

Research on sex identification of chicken eggs based on visible-near infrared hyperspectral technology

中文关键词:  可见-近红外高光谱  海兰褐鸡种蛋  性别判定  无损检测

英文关键词:visible-near infrared hyperspectral  Hylan brown chicken eggs  gender determination  nondestructive testing

中文摘要:

      目的 为实现鸡种蛋性别的无损检测, 提出了基于可见-近红外高光谱检测海兰褐鸡种蛋性别的方法。方法 通过分析种蛋0~14 d大头部位的400~1000 nm波段下的光谱, 建立基于偏最小二乘判别分析(partial least squares-discriminant analysis, PLS-DA)和支持向量机(support vector machine, SVM)的种蛋性别判别模型, 比较不同孵育天数下的模型判别率, 优选出最佳的检测天数; 通过分析4种不同的预处理算法, 选出最佳的鸡种蛋高光谱预处理方法, 最后构建基于全波段和特征波段光谱信息的判别模型, 并对结果进行比较。结果 基于PLS-DA和SVM的模型在第9 d的预测集结果达到最高, 分别为80.00%和82.50%。变量标准化(standard normalized variate, SNV)为最佳预处理方法; 全波段相对于连续投影算法(successive projections algorithm, SPA)、竞争性自适应重加权算法(competitive adapative reweighted sampling, CARS)选择特征波长的模型更优, 建模集、预测集准确率分别为90.00%和85.00%。结论 可见-近红外高光谱技术可以快速、较准确、无损检测海兰褐鸡种蛋性别, 该技术为褐壳鸡种蛋性别鉴定实现在线检测提供了一定的理论基础。

英文摘要:

      Objective To propose a method based on visible-near infrared hyperspectroscopy for determination the sex of Hylan brown chicken eggs, and realize the non-destructive detection of the sex of chicken eggs. Methods By analyzing the spectrum in the 400-1000 nm band of the large head of the eggs from 0 to 14 days, an egg sex discrimination model based on partial least squares-discriminant analysis (PLS-DA) and support vector machine (SVM) was established to compare model discrimination rate at different incubation days, the optimal detection days were selected; by analyzing four different preprocessing algorithms, the optimal hyperspectral preprocessing method for chicken embryos was selected, and finally the spectral information based on the full band and characteristic band was constructed, and the results were compared. Results It was found that the models based on PLS-DA and SVM achieved the highest results in the prediction set on the 9th day, with 80.00% and 82.50%, respectively. Standard normalized variate (SNV) was the best preprocessing method; compared with the continuous projection algorithm, successive projections algorithm (SPA) and the competitive adapative reweighted sampling (CARS), the full-band model was better than the model that selects the characteristic wavelength, and the accuracy rates of the modeling set and prediction set were 90.00% and 85.00%, respectively. Conclusion Visible-near infrared hyperspectral technology is able to detect the sex of Hylan brown chicken eggs quickly, accurately and non-destructively and this technology provides the theoretical basis for the online detection of brown-shell chicken eggs sex detection.