Research on Hyperspectral Images of Remote Sensing Based on Machine Learning Technology

This research focused on assessing the potential of hyperspectral remote sensing technology in the early identification of wheat leaf rust. The findings revealed that when hyperspectral remote sensing is integrated with machine learning techniques, it proves to be highly efficient in the early detection of wheat leaf rust. This combination not only enhances the accuracy of disease detection but also offers a promising approach for timely intervention in wheat crop management. Specifically, the Support Vector Machine (SVM) classification method, which utilized models built on 11 critical wavelengths (502, 466, 598, 718, 534, 766, 694, 650, 866, 602, and 858 nm), achieved data classification accuracy rates ranging from 97% to 100%. The analysis of VIS-NIR spectral data showed a significant correlation with hyperspectral remote sensing data within that range. However, direct comparison with GC-MS data was challenging because leaf pigments, mainly detected by VIS-NIR analysis, represent only a small part of the metabolome. Consequently, the early detection method for wheat leaf rust discussed in this study is based on classifying changes in pigment composition during disease progression. The effectiveness of this approach depends on the early identification of these pigment changes. These changes are effectively captured and verified by hyperspectral remote sensing and subsequently processed using SVMs. This approach holds significant promise for the accurate and timely detection of wheat leaf rust, facilitating early intervention measures and potentially reducing crop losses. By leveraging this advanced technology, farmers and agricultural professionals can implement more proactive disease management strategies, ultimately enhancing crop health and productivity.