Crowdsourcing Big Data Applications in Agriculture
Brianna B. Posadas
University of Florida
Big data has become an integral aspect of precision agriculture and modern farming in the United States. While farmers have been collecting data about various conditions of their farm for decades, new technologies have been able to store, analyze, and create new software from the data to help farmers make better predictions about their yields and better manage their farm. One of the challenges of using big data in agriculture is the dependence on people to create the ground truth data, or geophysical parameter data, to create the training data for precision agriculture machine learning algorithms. As it has become cheaper and easier to collect aerial images and other remote sensing data, it is still vital that someone physically inspect the area to identify the target, i.e. identify the vegetation, disease, or pest of interest. Ground truthing requires a lot of labor that farmers and researchers alone cannot satisfy. A solution to alleviating the labor shortage is to crowdsource the ground truthing. Other crowdsourcing techniques have been used to help train machine learning algorithms to identify plants and assist in discovering protein folding techniques. This dissertation will generate a prototype to 1) teach the user about the desired characteristics to be identified or “ground truthed” 2) direct the user to GPS coordinates 3) allow the user to make a real-world classification and report it and 4) allow for the user to verify classifications of other users to ensure data quality.