The chicken is easy to get spoilage during the storage and transportation. The hyperspectral imaging technology with the maps and spectral combination features was applied in this study, which could extract the spectral characteristics, texture and color characteristics to realize the rapid classification from the intrinsic quality and external qualities of the chicken. First of all, according to the physical and chemical analysis, 62 chicken breast samples were divided into four categories which are safe, edible, not recommended, and inedible. Secondly, using 42 labeled samples as the training set, texture, color, and spectral features were taken as the input of K-means-RBF neural network to identify the K-means initial classification center, train the RBF neural network, and construct the K-means-RBF chicken quality classification model. K-means-RBF chicken quality classification model was tested using the remaining 20 samples as test sets. The test results showed that the accuracy of the classification of 20 test samples was 100%. However, the classification accuracies of texture feature, color feature and texture color were 85%, 80%, and 95%, respectively. The chicken quality classification has successfully utilized the feature of hyperspectral imaging “map and spectral combination”, and the comprehensive detection of chicken has been successfully achieved. The results show that the K-means-RBF fusion method is effective in the hyperspectral data analysis, and the limitation of single feature in classification.