<p>This paper presents a fused framework composing three modules for the classification of crops focusing different feature sets and classification techniques. The first module is supervised machine learning model based on Random Under-Sampling (RUSBoost) classification algorithm implemented by taking the input of bio-spectral ensemble features and some specified spectral bands extracted from hyperspectral datasets. The second module is referred as a semi-supervised learning model based on Stacked Autoencoder (SAE) network. SAE is used to extract and classify endmembers, which are pure spectral signatures indicative of different crop types which enhance the network's ability to recognize subtle spectral variations between crops. The third module utilized a Convolutional Neural Network (CNN) that emphasizes the morphological, textural and color based characteristics of the crops. Finally, the outputs of these three modules are combined into a hybrid framework using an ensemble learning approach to produce more robust and accurate classification systems. With the integration of the contribution of each of the bio-spectral ensemble set, specified spectral bands, endmember features, morphological, texture, and color-based features in the hybrid framework, the overall classification accuracy has been substantially enhanced. This fused model have been tested and validated through five crop datasets i.e. Indian Pines (IP), Salinas (SA), WHU_Hi_LongKou (WHLK), WHU_Hi_HanChuan (WHHC) and WHU_Hi_ HongHu (WHHH). For these datasets, the proposed fused model obtained the highest classification accuracies of 96.83%, 99.71%, 99.87%, 98.09%, and 98.53% respectively. The experimental results show substantial improvements in classification accuracy, thus demonstrating the potential of the hybrid model for crop classification.</p>