Hybrid fusion of E-nose and computer vision using optimized deep learning and machine learning for robust plant leaf recognition

Pouya Bohlol, Mohammad Hasan Sabet Dizavandi, Syed Saeid Mohtasebi, Mahmoud Omid

Abstract The fusion multi-sensory system with optimized deep learning and machine learning algorithms appeared to synergize difficult paradigms in precision agriculture and boost recognition of various plant species. In this study, an electronic nose (E-nose) system with eight MOS sensors and a computer vision platform were designed for constructing a large-scale data set of 26 various species of plant leaves collected during 2 years under different environmental conditions. After preprocessing functions, the E-nose data set consisted of 15,600 data points, while two-phase data augmentation expanded the image data set to 156,000 samples. Multiple machine learning models, including MLPs and 7 machine learning algorithms, were executed and optimized based on the E-nose dataset, where KNN with 97% accuracy and 5 ms response time per sample showed the highest accuracy in recognizing leaves. In parallel, six deep learning and vision transform algorithms were implemented, and hyperparameters such as batch size, learning rate, optimizer, and image size were updated during training. InceptionV3 demonstrated superior performance compared to others, with 99% accuracy, 0.2 loss, and 8 ms response time. Finally, a fusion model of E-nose and computer vision-based deep and machine learning algorithms was developed for robust performance of recognizing 26 different species of leaves. The fusion model could recognize leaves with 99.8% accuracy with a 10-ms response time. The proposed hybrid approach highlights the potential of multi-modal fusion for reliable, fast, and scalable solutions in smart agriculture.