基于卷积神经网络与高光谱的鸡肉品质分类检测

高光谱成像技术是现代食品检测中的重要方法，根据其图、谱合一的特点，从鸡肉的高光谱数据中提取反映鸡肉内部品质的光谱数据和反映鸡肉外部特征的图像数据，对提取到的数据进行预处理，建立基于光谱和彩色图像的卷积神经网络（convolutional neural network，CNN）模型，对鸡肉的品质进行快速、无损检测。结果表明，基于光谱和图像的综合CNN模型的分类效果最好，其准确率和损失函数分别达93.58%和0.30，优于使用单一数据的CNN模型，证明综合使用鸡肉的内、外信息能够有效提高鸡肉品质检测精度。

Application of Convolutional Neural Network Combined with Hyperspectral Imaging in Chicken Quality Classification

Hyperspectral imaging is an important modern technique rnin food detection. This study presented the application of hyperspectral imaging to rapidly and nondestructively chicken quality. The hyperspectral data reflecting internal quality and external characteristics of chicken were extracted and they were preprocessed to establish a convolutional neural network (CNN) model based on the spectra and/or the color images. The results showed that the integrated CNN model based on both the spectra and the color images had the best performance for chicken quality classification with an accuracy and loss function of 93.58% and 0.30, respectively, demonstrating that the integrated use of internal quality and external characteristics of chicken can effectively improve the detection accuracy of chicken quality.