作者: Zhenjie Xiong , Da-Wen Sun , Hongbin Pu , Zhiwei Zhu , Man Luo
DOI: 10.1016/J.LWT.2014.10.021
关键词:
摘要: This study investigated the potential of visible and near infrared (Vis/NIR) hyperspectral imaging (HSI) to differentiate between free-range broiler chicken meats. 120 images fillets were acquired then calibrated for reflectance. Spectral data extracted from region interest (ROI), followed by multiple scatter correction (MSC) reduce noise. Successive projection algorithm (SPA) was used select optimal wavelengths full spectra. On other hand, principal component analysis (PCA) applied optimum characteristic images, first two (PC) selected because PC1 PC2 explained over 95% variances all Then, gray-level gradient co-occurrence matrix (GLGCM) implemented on extract 30 textural variables in total. Based fusion, classification models established, which radial basis function-support vector machine (RBF-SVM) model gave best results with high correct rate (CCR) 93.33% prediction samples, demonstrating that combining spectra texture effective differentiating