<p>Globally, food safety is a critical factor in the food supply chain process for public health concerns. The food contamination caused by pathogens and other physical hazards increases the complexity in supply chain process leads to mislabelling and adulteration. Additionally, the inconsistencies in different regions and development of antibiotic resistance in the bacteria and environmental cases complicate the process. The advancement of hyperspectral imaging and machine learning technique improves the traditional inspection methods to provide faster and accurate information about contamination in food. This paper developed an effective classification model with the feature extraction process to detect food contamination. The proposed model processes the hyperspectral images of the foods to eliminate the noises. With the processed images dual-tree cross-wavelet transformation (DT-CWT) is employed for the extraction of the features. With the extracted features Convolutional Neural Network (CNN) integrated with Multilayer Perceptron (MLP) is employed for the classification of the food contamination. Simulation analysis demonstrated that proposed model captures the spatial and spectral values of the food images in high dimensional hyperspectral data. The analysis is conducted with the 140 hyperspectral images with categories of Cedar Apple Rust, Apple Scab, Black Rot, and Healthy Leaves. The proposed classification model with MLP achieving an accuracy of 82.5%, while CNN models optimized with SGD and Adam reached 88.7% and 91.3% accuracy, respectively. The proposed CNN + MLP model outperformed these with an overall accuracy of 94.1%, precision of 92.6%, recall of 93.1%, and an F1-score of 92.8%. Additionally, at 50 epochs, the classification of Healthy Leaves achieved an impressive accuracy of 98.0%. The analysis demonstrated that proposed classification model with the cross-wavelet improves the accuracy with the effective food contamination detection.</p>