Numerical Modeling and Design of Machine Learning Based Paddy Leaf Disease Detection System for Agricultural Applications

In order to satisfy the insatiable need for ever more bountiful harvests on the global market, the majority of countries deploy cutting-edge technologies to increase agricultural output. Only the most cutting-edge technologies can ensure an appropriate pace of food production. Abiotic stress factors that can affect plants at any stage of development include insects, diseases, drought, nutrient deficiencies, and weeds. On the amount and quality of agricultural production, this has a minimal effect. Identification of plant diseases is therefore essential but challenging and complicated. Paddy leaves must thus be closely watched in order to assess their health and look for disease symptoms. The productivity and production of the post-harvest period are significantly impacted by these illnesses. To gauge the severity of plant disease in the past, only visual examination (bare eye observation) methods have been employed. The skill of the analyst doing this analysis is essential to the caliber of the outcomes. Due to the large growing area and need for ongoing human monitoring, visual crop inspection takes a long time. Therefore, a system is required to replace human inspection. In order to identify the kind and severity of plant disease, image processing techniques are used in agriculture. This dissertation goes into great length regarding the many ailments that may be detected in rice fields using image processing. Identification and classification of the four rice plant diseases bacterial blight, sheath rot, blast, and brown spot are important to enhance yield. The other communicable diseases, such as stem rot, leaf scald, red stripe, and false smut, are not discussed in this paper. Despite the increased accuracy they offer, the categorization and optimization strategies utilized in this work lead it to take longer than typical to finish. It was evident that employing SVM techniques enabled superior performance results, but at a cost of substantial effort. K-means clustering is used in this paper segmentation process, which makes figuring out the cluster size, or K-value, more challenging. This clustering method operates best when used with images that are comparable in size and brightness. However, when the images have complicated sizes and intensity values, clustering is not particularly effective