Satellite Image-Based Crop Disease Detection Using Deep Learning

Crop diseases are an issue for farmers. They can really hurt crops. Affect the food we eat. Small farmers in developing countries have it the worst because they do not have the money to find diseases early. Currently experts have to go into the fields to look at the plants for diseases. This way of doing things does not work well for farms. We used pictures from satellites and computer programs to make a system that finds crop diseases. We got information from Landsat satellites, which take pictures of crops and find signs that they're not healthy. Our system looks at this information. Points out fields that look different, from the others. Crop diseases are what our system is trying to find. A custom Convolutional Neural Network then takes over. It classifies each field as healthy or diseased. We trained this model using a dataset that includes satellite observations and information from field surveys. Crop diseases are an issue for farmers everywhere. Losing crops can have severe consequences. When a disease strikes, it affects food on our tables, the income farmers depend on and even the wider economy. If farmers do not identify diseases early, their choices become limited. They. Watch their harvests decline, or they apply pesticides sometimes at a great financial cost or in ways that harm the land and water. The real issue is timing. Pathogens quietly enter a field. Linger before anyone notices. Initially, a few plants get infected. It may be an area, nothing obvious. Fungi, bacteria and viruses start small. Then spread as conditions improve. The old ways of tracking disease do not work well. Countries agricultural extension services do not have a lot of money. They do not have enough staff. They focus on issues that already need to be addressed not on catching problems before they get worse. In places like Saharan Africa one extension worker may have to help thousands, tens of thousands, of farmers. In countries experts usually only look at high-value crops or check fields when it is really important. This means most fields are not checked for most of the season. Farmers, who may not know a lot about plant disease are often left to manage on their own with their crops. We used satellite images to create a system that identifies crop diseases. We also made an app that works on platforms. Farmers and agricultural officers can check crop health maps get warnings and find advice on diseases on their phones even in areas with bad internet. The app connects to a cloud system that collects and looks at satellite data then shows the results in a way. Our system is very good at identifying crop diseases with 90.2% accuracy for fields. It is good at finding diseases and telling us when it is right.We want to make the satellite data clearer make models that track disease over time and add features that work without internet and more language options, to the app for farmers to use with their crops.