INTELLIGENT UAV SCOUTING FOR FIELD CONDITION MONITORING

• Main Author: Seyyedhasani, Hasan
• Format: Others
• Published: UKnowledge 2018
• Subjects: Unmanned Aerial Vehicle; Scouting; Fuzzy Inference System; Average Sampling Probability; Total Scouting Luminance; Genetic Algorithm; Bioresource and Agricultural Engineering; Robotics
• Online Access: https://uknowledge.uky.edu/ece_etds/113; https://uknowledge.uky.edu/cgi/viewcontent.cgi?article=1119&context=ece_etds

Precision agriculture requires detailed and timely information about field condition. In less than the short flight time a UAV (Unmanned Aerial Vehicle) can provide, an entire field can be scanned at the highest allowed altitude. The resulting NDVI (Normalized Difference Vegetation Index) imagery can then be used to classify each point in the field using a FIS (Fuzzy Inference System). This identifies areas that are expected to be similar, but only closer inspection can quantify and diagnose crop properties. In the remaining flight time, the goal is to scout a set of representative points maximizing the quality of actionable information about the field condition. This quality is defined by two new metrics: the average sampling probability (ASP) and the total scouting luminance (TSL). In simulations, the scouting flight plan created using a GA (Genetic Algorithm) significantly outperformed plans created by grid sampling or human experts, obtaining over 99% ASP while improving TSL by an average of 285%.