Integration of convolutional neural networks and remote sensing imagery for automated identification and severity estimation of plant leaf diseases in large-scale farming

Speaking of increasing crop production and decreasing losses in commercial growing, the timely prediction and evaluation of the level of severity of plant leaf diseases can be named as a highly important element. This paper given in more detail illustrates more classes of leaf diseases through a combined approach that involves the use of CNNs and fine-dimensional remote sensing images. The diseases will automatically be identified and categorized with a massive number of leaves that will be scattered in big farming communities. This has been achieved by use of spatial-spectral feature extraction technique. Regulation of the kind of illness and the extent of severity can be done with the help of multi-stage CNN pipeline, which has been trained to make use of a dataset, which has been specifically annotated and which has varied lighting and field conditions. The areas that are calculated in order to determine the severity of the condition are ratios of areas of the lesions on the leaves, and the obtained results are made up as the geospatial maps thus allowing a high area to be analyzed. The module learning on ensemble can put together the responsibility of numerous CNN models and produce more precise results, a requirement that can be used to consolidate reliability. Under the cloud-based implementation, the option to receive real-time feedback as well as the ability to process the data of the agronomists becomes possible. The technology can be scaled, easily and flexibly accurate, and viable to utilize in precision agriculture particularly under resource-demanding situations, which offers an intelligent facilitation feature to the farmers and the agencies of agriculture. The technique is accurate up to 97.4% and like other methods, the accuracy is broad scale and as such, favorable to other methods of disease diagnosis.